preflight

Run a rigorous, multi-dimensional quality audit of a plan, requirements set, or any artifact — grounded in the actual codebase.

What it does

preflight is an adversarial review gate. It hunts for every ambiguity, contradiction, gap, and risk in an artifact, verifies every claim against the real code, scores findings on a 13-dimension scan, and presents each finding with options and a recommendation for you to decide.

It never fixes anything silently — it is a review protocol that finds and reports issues, applying changes only when you explicitly ask.

For high-stakes findings (CRITICAL/MAJOR), preflight hardens its recommendation with an independent challenger before presenting it, and closes consequential findings with a Confidence: / Hardening: line (see Concepts → Recommendation hardening).

When to use it

• After make_requirements or make_plan, before exec_plan.
• Any time you want a codebase-grounded audit of a plan, requirements, or an ad-hoc document.
• "Audit this plan", "review gate", "quality audit", "review my requirements before I build".

Trigger phrases

"preflight", "preflight <artifact>", "audit this plan", "audit these requirements", "quality audit", "review gate". Resume with --continue.

Worked example

preflight my-feature

Audits plans/my-feature/ against the real codebase, scores findings across 13 dimensions, and walks you through each one with a recommendation. You can also target a single document (preflight my-feature 03-api-design) or a requirements RD (preflight requirements RD-03).

Related skills

• make_plan / make_requirements — the artifacts preflight audits.
• exec_plan — run it after a clean preflight.
• upgrade_plan — if preflight shows the artifact is below current standards.